Chronic optical access through a polished and reinforced thinned skull.

Abstract

We present a method to form an optical window in the mouse skull that spans millimeters and is stable for months without causing brain inflammation. This enabled us to repeatedly image blood flow in cortical capillaries of awake mice and determine long-range correlations in speed. We also repeatedly imaged dendritic spines, microglia and angioarchitecture, as well as used illumination to drive motor output via optogenetics and induce microstrokes via photosensitizers.

Full text

Page 1

Chronic imaging and manipulation of cells and vessels through
a polished and reinforced thinned-skull
Patrick J. Drew1,2,3, Andy Y. Shih1, Jonathan D. Driscoll1, Per Magne Knutsen1, Pablo
Blinder1, Dimitrios Davalos4, Katerina Akassoglou4,5, Philbert S. Tsai1, and David
Kleinfeld1,6,7
1 Department of Physics, University of California at San Diego, CA
2 Center for Neural Engineering, Department of Engineering Science and Mechanics, The
Pennsylvania State University, University Park, PA
3 Department of Neurosurgery, The Pennsylvania State University, University Park, PA
4 Gladstone Institute of Neurological Disease, University of California, San Francisco, CA
5 Department of Neurology, University of California, San Francisco, CA
6 Graduate Program in Neurosciences, University of California at San Diego, CA
7 Center for Neural Circuits and Behavior, University of California at San Diego, CA
Abstract
We present a method to form an optical window in the mouse skull that spans millimeters and is
stable for months without inflammation of the brain. This enabled us to repeatedly image blood
flow in cortical capillaries of awake animals and determine long-range correlations in speed. We
further demonstrate repeated cortical imaging of dendritic spines, microglia, and
angioarchitecture, as well as illumination to drive motor output via optogenetics and induce
microstrokes via photosensitizers.
Keywords
Microscopy; Imaging; Neuroscience
Chronic observation and manipulation of cells in the cortex is integral to studies on neurons,
glia, and microvasculature. An increasingly popular application involves the use of in vivo
two-photon laser scanning microscopy (TPLSM) to image deep within mammalian cortex
through a window in the skull1. This may be formed by a craniotomy2, in which a section of
skull is removed and replaced with glass, or by thinning a section of the skull with a
microsurgical blade3. Both methods have been employed for the chronic observation of
Users may view, print, copy, download and text and data- mine the content in such documents, for the purposes of academic research,
subject always to the full Conditions of use: http://www.nature.com/authors/editorial_policies/license.html#terms
Correspondence: David Kleinfeld, Department of Physics, University of California, 9500 Gilman Drive, La Jolla, CA 92093-0374,
Phone: 858-822-0342, dk@physics.ucsd.edu.
Author contributions
The PoRTS window was conceived by PJD, AYS and PST, the imaging tools were developed by JDD and DK, the experiments were
designed by KA, DD, PJD, DK, AYS and PST and carried out by PJD, PMK, AYS and PST, data analysis was carried out by PB,
PJD, DK and PST, and the manuscript was written by PJD, DK and AYS.
Competing interests statement
The authors declare no competing financial interests.
NIH Public Access
Author Manuscript
Nat Methods. Author manuscript; available in PMC 2011 October 30.
Published in final edited form as:
Nat Methods. 2010 December ; 7(12): 981–984. doi:10.1038/nmeth.1530.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Page 2

fluorescently labeled neuronal processes3,4 and microvasculature5, yet have limitations. It
has been suggested that cranial windows activate microglia and astrocytes as part of an
inflammatory response6 that can alter neuronal physiology7 and pial blood vessels8,
although these complications may relate to surgical finesse9. The thinned skull procedure
avoids the activation of microglia and astrocytes6, yet access is limited to a very small area,
i.e., ~ 0.2 mm2 compared with 1 to 10mm2 for a craniotomy. Further, sequential imaging
sessions require re-thinning of the skull to counteract re-growth of the bone, which limits the
number of times that imaging can be performed and leaves the area under the thinned
portion of skull mechanically vulnerable3. Thus there is a need for a technique that provides
readily available optical access across a large area of the brain without the potential to
induce inflammation.
We introduce a polished and reinforced thinned-skull (PoRTS) procedure to create a large,
chronically stable window in the skull. We attempt to satisfy four goals with this method:
(1) Optical clarity, sufficient for imaging with TPLSM as well as optogenetic control,
without the need to intervene after the initial surgery; (2) Optical access across many square
millimeters; (3) Mechanical stability for imaging in awake animals; and (4) The absence of
an inflammatory response.
Our basic idea is to mechanically thin the skull3, polish the cut surface to improve flatness,
then use transparent cement and glass to form a clear, rigid window that is fused to the
thinned bone. The glass window is designed to stabilize the thinned and thus weakened
bone, and both the mechanical stability and the fusion of the window to the bone is reasoned
to inhibit regrowth of bone, thus preserving clarity. In detail, the skull is thinned with a high
speed burr to a final thickness of 10 to 15 μm, polished without direct pressure on the skull
with a slurry of grit in artificial cerebral spinal fluid (ACSF), and then fused to a cover glass
with a thin layer of cyanoacrylate cement (Fig. 1A).
The optical transfer function for two-photon excited fluorescence measured through the
PoRTS window is modestly degraded compared to that found using only a cover glass (Fig.
1B). The resolution varies across the window, presumably as a result of natural
inhomogeneities in bone, with a radial width of 0.7 to 1.0 μm (full width of half of the
integrated intensity) and an axial resolution of about 3 μm. Further, the TPLSM signal is
diminished by a factor of e–1.5 by the thinned bone (Supplementary Fig. 1); this is
equivalent to the loss in signal by imaging through 150 μm of brain tissue10. Despite such
caveats, we could repeatedly observe the dendritic spines of neurons that express yellow
fluorescent protein (YFP) both two days and several weeks after surgery (2 animals) (Fig.
1C). Second, we repeatedly imaged subsurface brain microvessels in animals in which the
plasma was stained with fluorophore-conjugated-dextran. Good visibility was obtained up to
250 μm below the pial surface for periods up to three months (35 animals) (Fig. 1D). In a
more stringent examination, we found that the pattern of subsurface microvascular
architecture was stable when the same group of microvessels was imaged over a period of
40 days through a PoRTS window (3 fields across 2 animals) (Fig. 1E). Consistent with the
long-term optical clarity of subsurface brain structures, the PoRTS window appears to
impede regrowth of the thinned skull as assayed by measuring the thickness of the skull over
time (Supplementary Fig. 2).
A critical biological assessment of the PoRTS window concerns the potential activation of
microglia; these cells change their morphology in response to insults to the brain11. We
utilized transgenic mice that express enhanced green fluorescent protein (EGFP) in
microglia, i.e., line Cx3cr1eGFP+/−. At two days after surgery, which corresponds to the time
when microglia are maximally active under a craniotomy6, we observe a diffuse, highly
branched morphology of microglia under the PoRTS window that is indicative of non-
Drew et al. Page 2
Nat Methods. Author manuscript; available in PMC 2011 October 30.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Page 3

activation (3 animals) (Supplementary Fig. 3). An essentially identical pattern is seen on the
control, contralateral hemisphere. Submicrometer microglial processes, at depths greater
than 100 μm below the surface, were readily imaged over a period of two months
(Supplementary Fig. 4). Repeated imaging of the same general area showed a continued
absence of activated microglia up to at least one month following installation of the PoRTS
window (5 animals at 2 days, 3 animals at 14 days, and 3 animals at 30 days)
(Supplementary Fig. 5).
A further assessment concerns reporters of brain trauma as evaluated within the area of
cortex just below the PoRTS window. First, we assessed the expression of glial fibrillary
acidic protein (GFAP), which is up-regulated in reactive astrocytes during brain injury12.
The pattern of immunostaining for GFAP was found to be weak and indistinguishable from
that seen on the control, contralateral hemisphere (3 animals) (Supplementary Fig. 6). This
implies a lack of astrocyte activation. Second, we exploited the past finding that cortical
injury can trigger remodeling and outgrowth of surface pial vasculature8. We observe that
the pial surface vasculature remains nearly unchanged (2 fields in each of 2 animals) with
the PoRTS window (Supplementary Fig. 7). In toto, the astrocyte, microglia and vascular
data show that the PoRTS window does not induce a classic inflammatory response in the
brain. By contrast, companion experiments reveal an increase in the density of brightly
immunostained GFAP-positive reactive astrocytes below a craniotomy (3 animals), as well
as an alteration of the topology of the vasculature (2 animals) that occurs primarily through
the growth of surface veins following a craniotomy (Supplementary Fig. 7).
An application that exploits the large area of cortex that is visible through the PoRTS
concerns the possibility of long-range correlations in vasomotion10, a 0.1 to 1 Hz oscillation
in vascular tone that is hypothesized to be part of the “default” state of brain activity13. We
made simultaneous measurements of the flow of red blood cells (RBCs) in pairs of
subsurface microvessels, defined as vessels with diameters less than 7.5 μm, in awake
animals. The RBCs can be measured in vessels over 100 μm below the pial surface and
separated by up to 500 μm (17 animals measured up to 60 days after implantation of the
window) (Fig. 2A-D). As a population, the baseline speed was heterogeneous (0.90 ±
0.61mm/sec; mean ± SD; 159 vessels) (Supplementary Fig. 8) and matches the values found
for anesthetized rat10. Of critical interest, low-frequency variations in blood flow were
strongly and significantly correlated out to 500 μm (Fig. 2E), the longest separation we
could probe. Coherence at the heart rate, which serves as a positive control, was strongest
between microvessels separated by less than 100 μm and remained significant out to 500
μm, Intermediate frequencies showed considerably less coherence. These data provides
direct evidence for high spatial coherence of the 0.1 to 1.0 Hz vasomotion in awake animals.
We now turn to the manipulation of brain activity. A first application concerns the excitation
of neurons expressing light-sensitive channels through the PoRTS window. We used a strain
of mice that express channelrhodopsin-2 (ChR2) in the layer 5 pyramidal neurons, i.e., thy–
1–YFP–-ChR2, and found that vibrissa movements were elicited by illumination of motor
cortex with blue light (2 awake animals) (Fig. 3A,B). This shows that focal excitation of
cells that express ChR2 is compatible with the PoRTS window. As a second application, we
show that we can induce microstrokes via photosensitizers through the PoRTS window. An
optically-mediated occlusion to a single penetrating arteriole, the vessel that sources blood
from the cortical surface to the paranchyma, is induced by focal illumination of the
photosensitizer Rose-Bengal, which circulates intravenously, with green laser light. This
leads to a localized clot (2 animals) (Fig. 3C,D). Occlusion of this single vessel is sufficient
to produce a significant infarct (Fig. 3E), as seen previously with a craniotomy14. Use of the
PoRTS window permits the effects of experimental stroke on cortical viability to be
characterized while avoiding the possibility of additional damage from the craniotomy
Drew et al.Page 3
Nat Methods. Author manuscript; available in PMC 2011 October 30.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Page 4

procedure. Overall, these examples show the utility of the PoRTS window for physiological
manipulation as well as imaging.
In summary, the PoRTS method enables high resolution chronic imaging and optical
manipulation through large windows into the brain for periods of at least three months. The
windows do not require regular maintenance, do not induce an apparent inflammatory
response nor lead to vascular rearrangement, and may be used with awake animals. The
downside of imaging through thinned bone is a loss in resolution, although this may be
ameliorated through the use of adaptive optics to correct for distortions of the wave-front of
the incident excitation beam. Lastly, the PoRTS window may be combined with
miniaturized, head mounted microscopes for chronic imaging in free-ranging animals15.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Acknowledgments
We thank Beth Friedman, Carlos Portera-Cailliau and Karel Svoboda for critical comments on an early version of
the manuscript, Michael Fuentes and Carl Petersen for advice on head fixation, Daniel N. Hill for discussions on
data analysis, Jennifer Lee and Katherine Yang for help with animal husbandry, Jeffrey D. Moore for discussions
on optical stimulation, and Steve Tayman for a gift of tin oxide. This work was supported by grants from the NIH
(EB003832, MH085499, NS059832 and RR021907 to DK and NS066361 to KA) and the Dana Program in Brain
and Immuno-imaging (to KA) and fellowships from the Israeli Science Foundation (to PB), the Canadian Institutes
of Health Research and American Heart Association (to AYS), the Human Frontiers Scientific Program (to PMK),
and the National Multiple Sclerosis Society (to DD).
Abbreviations
ACSF
Artificial cerebral spinal fluid
ChR2
Channelrhodopsin-2
eGFP
Enhanced green fluorescent protein
GFAP
Glial fibrillary acidic protein
PoRTS
Polished and reinforced thinned-skull
RBC
Red blood cell
TPLSM
Two photon laser scanning microscopy
YFP
Yellow fluorescent protein
References
1. Svoboda K, Denk W, Kleinfeld D, Tank DW. Nature. 1997; 385:161. [PubMed: 8990119]
2. Levasseur JE, Wei EP, Raper AJ, Kontos AA, Patterson JL. Stroke. 1975; 6:308. [PubMed:
1154467]
3. Grutzendler J, Kasthuri N, Gan WB. Nature. 2002; 420:812. [PubMed: 12490949]
4. Trachtenberg JT, Chen BE, Knott GW, Feng G, Sanes JR, Welker E, Svoboda K. Nature. 2002;
420:788. [PubMed: 12490942]
5. Brown CE, Li P, Boyd JD, Delaney KR, Murphy TH. Journal of Neuroscience. 2007; 27:4101.
[PubMed: 17428988]
6. Xu HT, Pan F, Yang G, Gan WB. Nature Neuroscience. 2007; 10:549.
7. Hauss-Wegrzyniak B, Lynch MA, Vraniak PD, Wenk GL. Experimental Neurology. 2002; 176:336.
[PubMed: 12359175]
Drew et al. Page 4
Nat Methods. Author manuscript; available in PMC 2011 October 30.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Page 5

8. Sohler TP, Lothrop GN, Forbes HS. Journal of Pharmacology and Experimental Therapeutics. 1941;
71:325.
9. Holtmaat A, Bonhoeffer T, Chow DK, Chuckowree J, De Paola V, Hofer SB, Hubener M, Keck T,
Knott G, Lee WC, Mostany R, Mrsic-Flogel TD, Nedivi E, Portera-Cailliau C, Svoboda K,
Trachtenberg JT, Wilbrecht L. Nature Protocols. 2009; 4:1128.
10. Kleinfeld D, Mitra PP, Helmchen F, Denk W. Proceedings of the National Academy of Sciences
USA. 1998; 95:15741.
11. Stoll G, Jander S. Progress in Neurobiology. 1999; 58:233. [PubMed: 10341362]
12. Pekny M, Nilsson M. Glia. 2005; 50:427. [PubMed: 15846805]
13. Fox MD, Raichle ME. Nature Reviews of Neuroscience. 2007; 8:700.
14. Blinder* P, Shih* AY, Rafie CA, Kleinfeld D. Proceedings of the National Academy of Sciences
USA. 2010; 107:12670.
15. Sawinski J, Wallace DJ, Greenberg DS, Grossmann S, Denk W, Kerr JND. Proceedings of the
National Academy of Sciences USA. 2009; 106:19557.
16. Jung S, Aliberti A, Graemmel P, Sunshine MJ, Kreutzberg GW, Sher A, Littman DR. Molecular
and Cellular Biology. 2000; 20:4106. [PubMed: 10805752]
17. Arenkiel BR, Peca J, Davison IG, Feliciano C, Deisseroth K, Augustine GJ, Ehlers MD, Feng G.
Neuron. 2007; 54:205. [PubMed: 17442243]
18. Frostig RD, Dory Y, Kwon MC, Masino SA. Proceedings of the National Academy of Sciences
USA. 1993; 90:9998.
19. Stosiek C, Garaschuk O, Holthoff K, Konnerth A. Proceedings of the National Academy of
Sciences USA. 2003; 100:7319.
20. Kleinfeld D, Delaney KR. Journal of Comparative Neurology. 1996; 375:89. [PubMed: 8913895]
21. Denk W, Delaney KR, Kleinfeld D, Strowbridge B, Tank DW, Yuste R. Journal of Neuroscience
Methods. 1994; 54:151. [PubMed: 7869748]
22. Tsai, PS.; Kleinfeld, D. Methods for In Vivo Optical Imaging, 2nd edition. Frostig, RD., editor.
CRC Press; Boca Raton: 2009. p. 59
23. Tsai PS, Friedman B, Ifarraguerri AI, Thompson BD, Lev-Ram V, Schaffer CB, Xiong Q, Tsien
RY, Squier JA, Kleinfeld D. Neuron. 2003; 39:27. [PubMed: 12848930]
24. Nguyen, Q-T.; Dolnick, EM.; Driscoll, J.; Kleinfeld, D. Methods for In Vivo Optical Imaging. 2.
Frostig, RD., editor. CRC Press; Boca Raton: 2009. p. 117
25. Nguyen QT, Tsai PS, Kleinfeld D. Journal of Neuroscience Methods. 2006; 156:351. [PubMed:
16621010]
26. Shih AY, Friedman B, Drew PJ, Tsai PS, Lyden PD, Kleinfeld D. Journal of Cerebral Blood Flow
and Metabolism. 2009; 29:738. [PubMed: 19174826]
27. Shih, AY.; Nishimura, N.; Nguyen, J.; Friedman, B.; Lyden, PD.; Schaffer, CB.; Kleinfeld, D.
Imaging in Neuroscience and Development. Yuste, R., editor. Vol. 2. Cold Spring Harbor
Laboratory Press; New York: 2009. in press
28. Valmianski I, Shih AY, Driscoll J, Matthews DM, Freund Y, Kleinfeld D. Journal of
Neurophysiology. 2010; 104:1803. [PubMed: 20610792]
29. Drew PJ, Blinder P, Cauwenberghs G, Shih AY, Kleinfeld D. Journal of Computational
Neuroscience. 2010; 29:5. [PubMed: 19459038]
30. McMenamin PG. Journal of Comparitive Neurology. 1999; 405:553.
31. Ferezou I, Haiss F, Gentet LJ, Aronoff R, Weber B, Petersen CCH. Neuron. 2007; 56:907.
[PubMed: 18054865]
32. Knutsen PM, Derdikman D, Ahissar E. Journal of Neurophysiology. 2005; 93:2294. [PubMed:
15563552]
33. Nishimura N, Schaffer CB, Friedman B, Lyden PD, Kleinfeld D. Proceedings of the National
Academy of Sciences USA. 2007; 104:365.
34. Born, M.; Wolf, E. Principles of Optics: Electromagnetic Theory of Propagation Interference and
Diffraction of Light. 6. Pergamon Press; Oxford: 1980.
Drew et al.Page 5
Nat Methods. Author manuscript; available in PMC 2011 October 30.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Page 6

Figure 1. PoRTS window procedure, optical properties, and basic capabilities
(A) Schematic of the window. The thinned and polished skull is protected with a cover glass
attached with cyanoacrylate cement. Dental cement is used to seal the edge of the cover
glass and provide support for the meniscus. The thinned area is typically 1 – 2 mm on edge.
(B) Point spread functions for TPLSM determined by imaging 0.2 μm fluorescent beads
were embedded in 1 % (w/v) agarose and imaged with a 40x, 0.8 NA water dipping
objective. The axial depth profile shown the integrated radial intensity along the optical axis,
i.e., z-axis, and the radial depth profile shows the integrated axial intensity along the radial
axis. The reported Δz and Δr are the full widths that encompass half of the integrated
intensity. The top left panel shows a bead imaged through only water along width depth
profiles. The top right panel shows a bead imaged under a #0 cover glass (black line) along
with depth profiles, along with the profiles for a #1.5 cover glass (grey line). The two
bottom panels show separate examples of beads imaged through an excised PoRTS window.
(C) Dendrites and spines of thy1-YFP expressing neurons two days (left panel) and 30 days
(right panel) after PoRTS window implantation. Images were taken 30 μm below the
surface. Reduction in the brightness of the YFP over time necessitated averaging of five
frames for images taken one month after implantation; the dwell time was 6 μs per pixel and
the average incident power was 35 – 70 mW. Representative image from a set of two mice
(D) Maximum z-axis projections across 65 μm of fluorescein-conjugated-dextran filled
vasculature through a PoRTS window 90 days after surgery. Each image is the average of
six frames, the z-step was 2 μm, the dwell time was 3 μs per pixel and the average power
was 25 – 120 mW. Representative images from a pool of 35 mice (E) Subsurface
microvessels are stable under the PoRTS window for over a month after implantation.
Images show maximum projections across 20 – 70 μm below the surface using an average of
5 frames, a dwell time of 6 μs per pixel, and an average power of 30 – 45 mW.
Representative images from a set of two mice.
Drew et al. Page 6
Nat Methods. Author manuscript; available in PMC 2011 October 30.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Page 7

Figure 2. Long-range coherence of the velocity of RBC flow in capillaries in the cortex of awake
head-fixed mice through a PoRTS window
Example data in panels A to D are taken from the same animal using a 10x,0.3 NA lens and
averaging over 5 – 10 frames (A) Top panel is the maximal z projection over 90 μm of
images of fluorescein-conjugated-dextran labeled vasculature, based on frames separated by
10 μm, two days after surgery. Bottom panel is an image from a single plane in the same
animal, but a different location, 51 days after surgery. The colored lines show the scan path
of the laser focus: green and orange are constant velocity segments along capillaries and
purple are minimum time segments between capillaries. (B) Space-time plots of one
segment of line scan data for each of two capillaries, with the calculated instantaneous
velocity for the entire 300 s run shown below. (C) Top panel shows power spectra for the
two velocity traces in panel B; 0.083 Hz bandwidth. A rhythmic component at ~ 10 Hz
results from the heartbeat. Bottom panel is the magnitude of the spectral coherence between
the velocities of the two capillaries as a function of frequency; 0.1 Hz bandwidth. Line
denotes a significance of p < 0.004, which corresponds to the inverse of twice the number of
degrees of freedom. Note the strong coherence at the heart rate and at the 0.1 – 1 Hz
vasomotor frequencies. (D) Examples of spontaneous velocity coherences between
capillaries obtained various times after surgery; 0.1 Hz bandwidth. Panels A-D show
representative data from a pool of 17 animals monitored up to 60 days after implantation (E)
The coherence as a function of distance (50 vessel pairs in 9 mice). Left panel is the mean
coherence in the 0.1 – 1.0 Hz vasomotor band. Center panel is the coherence at the heart-
rate, which serves as a positive control for the maximum possible coherence. Right panel is
the coherence in the 2 to 6Hz band, which serves as the null hypothesis. Colored lines
correspond to fits with exponential functions to the corresponding spectral band.
Drew et al.Page 7
Nat Methods. Author manuscript; available in PMC 2011 October 30.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript

Page 8

Figure 3. Examples of cortical physiology evoked through the PoRTS window
(A) Sample frame from video of awake, head-fixed ChR2 carrying mouse with a PoRTS
window fabricated over vibrissa motor cortex 150 days earlier. Green line marks the
position of the caudal vibrissa that is tracked over time. (B) Rostral-caudal motion of a
caudal vibrissa in response to pulses of 467 nm light (5 Hz, 100 ms pulse duration; blue
lines) from a light emitting diode (LED). Larger angles indicate retraction. A-B
representative data from a set of two animals. (C and D) Maximal projection (x-y and y-z)
through a 200 μm depth of a Cx3cr1eGFP+/− mouse cortex before (panel C) and 100 minutes
after (panel D) occlusion of a single penetrating arteriole using targeted optical activation of
Rose Bengal (yellow arrow points to clot). Images are averages of four frames with a dwell
time of 3 μs per pixel and 1 μm steps along the z-axis. Note the change in microglia
morphology to an amoeboid shape around the occluded arteriole and the interruption of flow
as evidenced by the absence of Texas Red in the penetrating vessel highlighted in the y – z
projection in the bottom pannels. The occlusion was made one day after implantation of the
window. (E) The extent of the infarct, for the same animal as in panels C and D, visualized
two days after the optically generated stroke. Note the invasion of eGFP labeled microglia
into the cyst. The data in panels C to E are representative images from a set of two mice.
Drew et al.Page 8
Nat Methods. Author manuscript; available in PMC 2011 October 30.
NIH-PA Author Manuscript
NIH-PA Author Manuscript
NIH-PA Author Manuscript